Locomotor patterns in cerebellar ataxia
12
Martino G1,2, Ivanenko YP2, Serrao M3,4, Ranavolo A5, d’Avella A2, Draicchio F5, Conte C3,6, Casali
3
C4, Lacquaniti F1,2,7
4 5
1 Centre of Space Bio-medicine, University of Rome Tor Vergata, Rome, Italy
6
2 Laboratory of Neuromotor Physiology, IRCCS Santa Lucia Foundation, Rome, Italy
7
3 Rehabilitation Centre Policlinico Italia, Rome, Italy
8
4 Department of Medical and Surgical Sciences and Biotechnologies, Sapienza University of
9
Rome, Latina, Italy 10
5 INAIL, Department of Occupational Medicine, Monte Porzio Catone, Rome, Italy
11
6 IRCCS C. Mondino, Department of Neuroscience, University of Pavia, Pavia, Italy
12
7 Department of Systems Medicine, University of Rome Tor Vergata, Rome, Italy
13 14 15 16 17 18 19 20 21 22 Correspondence to: 23 Dr Giovanni Martino 24
Laboratory of Neuromotor Physiology, IRCCS Fondazione Santa Lucia, 306 via Ardeatina, 00179 25 Rome, Italy 26 tel. ++39 06.51.50.14.75 27 fax. ++39 06.51.50.14.82 28 e-mail: [email protected] 29
Articles in PresS. J Neurophysiol (September 3, 2014). doi:10.1152/jn.00275.2014
Abstract 30
31
Several studies demonstrated how cerebellar ataxia (CA) affects gait, resulting in deficits in 32
multi-joint coordination and stability. Nevertheless, how lesions of cerebellum influence the 33
locomotor muscle pattern generation is still unclear. To better understand the effects of CA on 34
locomotor output, here we investigated the idiosyncratic features of the spatiotemporal structure of 35
leg muscle activity and impairments in the biomechanics of CA gait. To this end, we recorded the 36
electromyographic (EMG) activity of 12 unilateral lower limb muscles and analyzed kinematic and 37
kinetic parameters of 19 ataxic patients and 20 age-matched healthy subjects during overground 38
walking. Neuromuscular control of gait in CA was characterized by a considerable widening of 39
EMG bursts and significant temporal shifts in the center of activity due to overall enhanced muscle 40
activation between late swing and mid-stance. Patients also demonstrated significant changes in 41
the intersegmental coordination, an abnormal transient in the vertical ground reaction force and 42
instability of limb loading at heel strike. The observed abnormalities in EMG patterns and foot 43
loading correlated with the severity of pathology (clinical ataxia scale, ICARS) and the changes in 44
the biomechanical output. The findings provide new insights into the physiological role of 45
cerebellum in optimizing the duration of muscle activity bursts and the control of appropriate foot 46
loading during locomotion. 47
48
Keywords: Cerebellar ataxia, gait adaptation, muscle activation patterns, central pattern 49
generator, limb loading. 50
Introduction 51
52
Cerebellum is known to play a critical role in the production of locomotor behavior (Lennard 53
and Stein, 1977; Grillner et al., 1995; Roberts et al., 1995; Rossignol et al., 1998). Several 54
physiological studies on animals, in fact, have described how lesions at different regions of 55
cerebellum are responsible for different deficits in locomotion (Thach and Bastian, 2004). Medial 56
cerebellar region plays a primary role in static and dynamic balance control and in modulating the 57
rhythmic flexor and extensor muscle activity. The intermediate and lateral cerebellar regions 58
appears to be more important for directing limb placement and fine adjustments to the normal 59
locomotor pattern in novel or complex circumstances or when strong visual guidance is required 60
(Morton and Bastian, 2007). 61
In humans, the gait impairment characterizing cerebellar ataxia (CA) is often clinically 62
described as “drunken gait”, as the clinical features typically observed include a widened base of 63
support and an irregular gait pattern (Holmes, 1939). Several studies investigated the 64
biomechanical characteristics of patients with CA, finding these to consist of decreases in step 65
length, gait speed and ankle torques, increased step width, impaired inter-joint coordination and 66
marked variability of all global and segmental gait parameter values (Palliyath et al., 1998; Mitoma 67
et al., 2000; Earhart and Bastian, 2001; Stolze et al., 2002; Morton and Bastian, 2003; Ilg et al., 68
2007; Serrao et al., 2012; Wuehr et al., 2013). All these gait abnormalities, reflecting a lack of limb 69
coordination and impaired balance, greatly restrict these patients in their daily life activities and 70
predispose them to falls (Van de Warrenburg et al., 2005b; Nardone and Schieppati, 2010). 71
No studies have yet been performed to provide a detailed analysis of muscle activity 72
patterns during locomotion in patients with CA. Furthermore, no information on the relationship 73
between the observed kinematic and kinetic abnormalities and the related muscle activity have 74
been provided so far. Therefore, the specific contribution of the cerebellum to the production of 75
locomotor muscle pattern behaviors in humans is still unclear. Impaired processing of sensorimotor 76
information in the cerebellum about foot kinematics and kinetics (Bosco et al., 2006) may also 77
disturb the limb loading and placement. Previous findings suggest that the cerebellum helps in 78
modulating sensorimotor interactions, integrates both feedforward and feedback control of balance, 79
and plays a functional role in motor learning and adaptation (Horak and Diener, 1994; Morton and 80
Bastian, 2004; Konczak and Timmann, 2007; Bastian, 2011; Goodworth et al., 2012; Ilg and 81
Timmann, 2013). Thus, lesions of the cerebellum may induce abnormalities in the spatial and 82
temporal pattern of muscle activation resulting in specific gait impairments. 83
The aim of this study was to provide a wider characterization of ataxic gait by exploring the 84
muscle activation patterns of patients affected by CA, and correlating them with the kinematics, 85
kinetics and the degree of severity of the pathology. The first objective was to test the hypothesis 86
that subjects with cerebellar damage show abnormalities in switching and scaling individual 87
muscles resulting in prolonged activity, as it occurs during upper limb movements in CA (Hallett et 88
al., 1975) or during early development of locomotion (Dominici et al., 2011) when the cerebellum is 89
still immature (Vasudevan et al., 2011). As a secondary objective, we studied the kinematic and 90
kinetic behavior of whole lower limb by analyzing the multiple joint coordination as well as the 91
ground reaction force during loading response. Particularly, we sought to investigate whether the 92
loading response in CA is impaired during weight acceptance, representing the most demanding 93
task to guarantee the initial limb stability and the preservation of progression (Perry, 1992). To 94
assess the coordination deficits, we used the methods developed earlier for normal gait (Borghese 95
et al., 1996; Lacquaniti et al., 2002) and we expected that applying this analysis technique to CA 96
patients may highlight specific alterations in the planar covariation of limb segment elevation 97
angles (Dominici et al., 2010; MacLellan et al., 2011; Leurs et al., 2012).To this end we 98
investigated a sample of patients with primary degenerative pancerebellar diseases. These 99
cerebellar disorders may represent an appropriate model to investigate the role of cerebellum as a 100
whole in locomotor pattern generation. 101
Materials and methods 102
103
Participants 104
Nineteen patients (5 females and 14 males; age range 32-65 yrs, weight 68±8 kg [mean ± 105
SD], leg length 0.78±0.06 m) affected by inherited CA, and twenty age-matched healthy subjects 106
HS (7 females and 13 males; age range 34-70 yrs, weight 70±14 kg, leg length 0.80±0.05 m) were 107
studied. The characteristics of patients are described in Table 1. Eleven patients had a diagnosis 108
of autosomal dominant ataxia (spinocerebellar ataxia, 7 pts with SCA1, 4 pts with SCA2), while the 109
other 8 had sporadic adult onset ataxia of unknown etiology (SAOA). Even if extracerebellar 110
involvement is common in both SCA1 and 2, none of our patients was found to have clinically 111
significant signs other than cerebellar ones. In particular, they did not show extrapyramidal or 112
pyramidal signs, nor signs of peripheral nerve or muscle deficits, which tend to be overt over the 113
course of the disease. All patients were at a relative early stage of disease as demonstrated by the 114
low International Cooperative Ataxia Rating Scale (ICARS) (Table 1), so that within the limits of 115
clinical ascertainment methods they can be regarded as relatively “pure” cerebellar patients. All 116
patients underwent a complete neurological assessment which included: (i) cognitive evaluation 117
(MMSE, mini-mental state scale); (ii) cranial nerves evaluation; (iii) muscle tone evaluation; (iv) 118
muscle strength evaluation; (iv) joint coordination evaluation; (v) sensory examination; (vi) tendon 119
reflex elicitation; (vii) disease severity measured by ICARS (Trouillas et al., 1997). Particularly, 120
sensation was tested clinically for light touch, pain, joint position and vibration, starting from the 121
toes and moving proximally; touch was tested by a wisp of cotton, pain by a sharp pin, vibration by 122
a 128-Hz tuning fork; proprioception was investigated by asking five times the blinded patient to 123
describe the position of the second toe and the ankle, which were passively moved upward or 124
downward by the examiner, avoiding end-of-range-of-motion position (DeMyer, 2003). In each of 125
these three sensory tests, we assessed whether sensation was normal, reduced or absent, 126
specifying the region of the body. All the patients were evaluated by two experienced neurologists 127
(CC and MS). 128
No patient had any kind of visual impairment, in particular no optic atrophy or retinitis 129
pigmentosa were revealed. On the other hand, almost all patients had oculomotor abnormalities 130
such as gaze nystagmus or square waves during pursuit movements, which are common in 131
cerebellar disorders with no obvious impairment of visual acuity. No patient showed clinical 132
features of spasticity, strength deficit, sensory deficit, and/or cognitive impairment (MMSE >26). 133
Furthermore, no relevant inter-limb asymmetries in terms of dysmetria, asynergia and hypotonia 134
and limb kinetic ICARS scores were found between right and left side. All patients showed (at MRI) 135
pancerebellar degenerations with significant atrophy of the cerebellar vermis. Patients enrolled in 136
our study were undergoing physical therapy, which included upper and lower limb exercises, 137
balance and gait training. All participants were capable of walking independently on a level surface, 138
they provided informed written consent prior to taking part in the study, which complied with the 139
Helsinki Declaration and had local ethics committee approval. 140
141
Procedures and data recording 142
Subjects were asked to walk barefoot along a walkway, approximately 7 m length, while 143
looking forward. They walked at comfortable self-selected speeds, but were encouraged to walk 144
also at the fastest speed at which they still felt safe, resulting in a range of different speeds across 145
the recorded trials. Given that typical walking speeds were on the slow side in patients, we 146
instructed the healthy control subjects to also walk barefoot at low comfortable speeds (~30-50% 147
slower than self-selected speed), in order to roughly match the walking speed in the two groups of 148
subjects. Before the recording session, subjects practiced for a few minutes to familiarize with the 149
procedure. To ensure safe walking conditions, an assistant walked alongside the patients during 150
the trials when necessary. In order to avoid muscle fatigue, groups of three trials were separated 151
by 1-min rest periods. At least 15 trials were recorded for each subject (≥10 trials at self-selected 152
speed and ≥5 trials at fast speed in patients or slow speed in controls) and the strides related to 153
gait initiation and termination were discarded so that each trial included from 1 to 3 consecutive 154
gait cycles. Only strides whose speed fell within the range of 3-4.5 km/h (since most trials were 155
performed in this range) were retained here for further analysis. Therefore the number of strides 156
analyzed in each subject (at matched walking speeds, 3-4.5 km/h) was on average 12.8±4.6 for 157
CA patients and 12.7±3.6 for control subjects, while the number of excluded strides was on 158
average 18.4±11.7 for CA patients and 10.4±3.8 for control subjects. 159
Kinematic data were recorded bilaterally at 300 Hz using an optoelectronic motion analysis 160
system (SMART-D System, BTS, Milan, Italy) consisting of 8 infrared cameras spaced around the 161
walkway. Twenty-two retro-reflective spherical markers (15 mm in diameter) were attached on 162
anatomical landmarks according to Davis et al. (1991). Anthropometric measurements were taken 163
on each subject. These included the mass and height of the subject and the length of the main 164
segments of the body according to the Winter’s method (Winter, 1979). Ground reaction forces 165
were recorded at 1200 Hz by means of two force platforms (0.6·m×0.4·m; Kistler 9286B, 166
Winterthur, Switzerland), placed at the center of the walkway, attached to each other in the 167
longitudinal direction but displaced by 0.2·m in the lateral direction. The EMG data were recorded 168
at 1000 Hz using a wireless system (FreeEMG300 System, BTS, Milan, Italy). Bipolar Ag-AgCl 169
surface electrodes were used to record EMG activity from 12 muscles simultaneously on the right 170
side of the body in each subject: tibialis anterior (TA); gastrocnemius lateralis (LG); gastrocnemius 171
medialis (MG); soleus (SOL); peroneus longus (PL); vastus lateralis (VL); vastus medialis (VM); 172
rectus femoris (RF); biceps femoris (BF); semitendinosus (ST); tensor fascia latae (TFL); gluteus 173
medium (GM). Innervation zones and tendon regions were identified using multi-channel high-174
density EMG recordings (Barbero et al., 2012) and SENIAM guidelines (Hermens et al., 1999) to 175
ensure correct placement of EMG electrodes. Acquisition of the EMG, kinematic, and kinetic data 176 was synchronized. 177 178 Data Analysis 179
GAIT CYCLE DEFINITION. Gait cycle was defined as the time between two successive foot 180
contacts of the same leg and foot strike and lift-off events were determined by maximum and 181
minimum excursions of the limb angle (Borghese et al., 1996; Vasudevan et al., 2011), defined as 182
the angle between the vertical axis and the limb segment (from the greater trochanter to lateral 183
malleolus) projected on the sagittal plane. When subjects stepped on the force platforms, these 184
kinematic criteria were verified by comparison with foot strike and lift-off measured from a threshold 185
crossing event in the vertical force (7% of body weight). In general, the difference between the time 186
events measured from kinematics and kinetics was no more than 3%. Nevertheless, since the 187
kinematic criterion produced a small error in the identification of stance onset, for averaging and 188
assessing the vertical ground reaction forces when subjects stepped on the force plate, we 189
identified the foot strike from the kinetic data. 190
191
KINEMATIC DATA PROCESSING. The following general gait parameters were calculated for
192
each subject: walking speed, cycle duration, relative stance duration, stride length and stride width. 193
The stride length and width were normalized to the limb length (thigh+shank) of each subject. We 194
computed both the anatomical joint angles and the elevation angles of the limb segments relative 195
to the vertical for the right lower limb (Borghese et al., 1996), as well as the pitch and roll angles for 196
the trunk. From these variables, we derived the range of angular motion (RoM). The kinematic data 197
were time interpolated over individual gait cycles to fit a normalized 200-point time base. We also 198
assessed the inter-stride angular variability by calculating the mean standard deviation (SD) of the 199
joint and trunk orientation angles. 200
Most coordination analyses of gait in CA (Earhart and Bastian, 2001; Stolze et al., 2002; 201
Morton and Bastian, 2003; Ilg et al., 2007) involved the use of angle–angle plots and thus 202
examined movement at two joints at a time. We used a more advanced analysis technique, termed 203
the planar law of intersegmental coordination, which allows to examine movement coordination at 204
the thigh, shank and foot segments simultaneously (Borghese et al., 1996; Lacquaniti et al., 2002). 205
Briefly, the temporal changes of the elevation angles at the thigh, shank, and foot covary during 206
walking. When these angles are plotted in three dimensions (3D), they describe a loop that can be 207
least-squares fitted to a plane over each gait cycle (Borghese et al., 1996). A principal component 208
analysis (PCA) was applied to the group of three segment elevation angle trajectories to determine 209
covariance loop planarity, width and orientation. To this end, we computed the covariance matrix of 210
the ensemble of time-varying elevation angles over each gait cycle. The first two eigenvectors 211
and lie on the best-fitting plane of angular covariation and the third eigenvector ( ) is the 212
normal to the plane and thus defines the plane orientation. The planarity of the trajectories was 213
quantified by the percentage of total variation ( ) accounted for by the third eigenvector of the 214
data covariance matrix (for ideal planarity 0%). Covariance loop width was determined using 215
the percent variance ( ) explained by the second eigenvector since it is oriented in the 216
direction of the minor axis of the loop formed by the elevation angles (if is small, the thigh-217
shank-foot loop tends to be a line, Ivanenko et al., 2008). Covariance plane orientation was 218
quantified using the direction cosine between the third principal axis and the positive semi-axis of 219
the thigh segment ( ), which was found to vary depending on walking conditions (Bianchi et al., 220
1998; Ivanenko et al., 2008) or gait pathology (Grasso et al., 1999, 2004; MacLellan et al., 2011; 221
Leurs et al., 2012). For each subject, the parameters of planar covariation ( , and ) were 222
averaged across strides. 223
224
GROUND REACTION FORCES. The steps in which only the right foot stepped onto one of the
225
force plates were analyzed. The vertical ground reaction force (GRF) was calculated and 226
normalized to the body mass (Winter, 1991). In addition, because the lower limb can undergo an 227
impulsive load, the heel strike transient (Verdini et al., 2006) during the weight-acceptance period 228
was evaluated by calculating the peak-to-peak change between the transient maximum and the 229
following local minimum in the GRF ( , Fig. 4A). If this transient was absent, the change was 230
considered equal to zero. Since the transient may also be related to limb instability or to small 231
oscillations of the limb during the heel strike, we evaluated the corresponding kinematics 232
correlates: the peak-to-peak change between maximum and minimum values of the vertical 233
velocity of three markers placed on the greater trochanter (hip), lateral femur epicondyle (knee) 234
and lateral malleolus (ankle) during the 0-10% interval of the stance phase ( , Fig. 4B). 235
236
ASSESSMENT OF EMGS. The raw EMG signals were band-pass filtered using a zero-lag 237
third-order Butterworth filter (20-450 Hz), rectified and low-pass filtered with a zero-lag fourth-order 238
Butterworth filter (10 Hz). The time scale was normalized by interpolating individual gait cycles over 239
200 points. For each individual, the EMG signal from each muscle was normalized to its peak 240
value across all trials. 241
To characterize differences in the amplitude and timing of EMG activity between CA and 242
HS groups, we computed the following parameters: mean and maximum EMG activity (in µV), 243
antagonist co-activation index ( ), center of activity ( ), and full width at half maximum 244
( ). EMG parameters were calculated over individual strides and then averaged across 245
cycles. 246
The was assessed between the thigh (mean activity of quadriceps RF-VL-VM vs 247
hamstring BF-ST) and calf (mean activity of triceps MG-LG vs TA) antagonistic muscle groups 248
using the following formula (Rudolph et al., 2000; Mari et al., 2014): 249
EMGH j EMGL j /2 EMGL j EMG⁄ H j
200 1
where EMGH and EMGL represent the highest and the lowest activity between the antagonist 250
muscle pairs. In order to have a global measure of the co-activity level, the was then averaged 251
over the entire gait cycle ( 1: 200). This method provided a sample-by-sample estimate of the 252
relative activation of the pair of muscles as well as the magnitude of the co-contraction over the 253
entire cycle. Using this equation, high co-contraction values represent a high level of activation of 254
both muscles across a large time interval, whereas low co-contraction values indicate either low 255
level activation of both muscles, or a high level activation of one muscle along with low level 256
activation of the other muscle in the pair (Rudolph et al., 2000). 257
The during the gait cycle was calculated using circular statistics (Batschelet, 1981) and 258
plotted in polar coordinates (polar direction denoted the phase of the gait cycle, with angle that 259
varies from 0 to 360°). The of the EMG waveform was calculated as the angle of the vector 260
(first trigonometric moment) which points to the center of mass of that circular distribution using the 261
following formulas: 262
sin 3 tan ⁄ 4
The was chosen because it was impractical to reliably identify a single peak of activity in the 263
majority of muscles, especially in pathological subjects. It can only be considered as a qualitative 264
parameter, because averaging between distinct foci of activity may lead to misleading activity in 265
the intermediate zone. Nevertheless, it can be helpful to understand if the distribution of muscular 266
activity remains unaltered across different groups and muscles. 267
The for each EMG waveform was calculated as the sum of the durations of the 268
intervals in which the EMG activity (after subtracting the minimum throughout the gait cycle) 269
exceeded the half of its maximum. 270
271
Statistics 272
Between groups differences in the spatiotemporal gait parameters, intersegmental 273
coordination, inter-stride variability and were assessed by performing unpaired two-sample 274
t-tests. The analysis of was performed using the Watson-Williams test for circular data 275
(Watson and Williams, 1956). The correlation between kinematics, kinetics, muscle activation 276
patterns and clinical scores was performed using Spearman’s rank correlation coefficient. The 277
correlation coefficients used in the regression plots were corrected for multiple samples from the 278
same participants. Descriptive statistics included means ± SD, and significance level was set at 279
0.05. All statistical analysis were performed using Statistica (v7.0) and custom software written 280
in Matlab (v8.1). 281
Results 282
283
General gait parameters and kinematics 284
At matched walking speeds, cerebellar patients showed a significant increase in the stride 285
width, reduction in the cycle duration and stride length in comparison with healthy controls (Fig. 286
1A). Instead, the relative stance duration was not significantly different in the two groups. These 287
results are consistent with previous studies (Palliyath et al., 1998; Mitoma et al., 2000; Serrao et 288
al., 2012). Figure 1B shows the ensemble-averaged kinematic patterns. The time course of 289
changes of hip and knee joint angles of CA patients was very similar to that of HS. In contrast, a 290
substantial reduction of the ankle joint excursion ( 0.00004) and consistently larger oscillations 291
in the trunk roll and pitch angles ( 0.0009) were observed in CA relative to HS (Fig. 1C). 292
Figure 2A illustrates the stride-averaged (±SD) thigh, shank and foot elevation angles and 293
corresponding gait loops plotted in 3D for one representative HS (left) and one CA patient (right). 294
In both groups, the temporal changes of the elevation angles covary during walking, describing a 295
characteristic loop over each stride that is best-fitted by a plane ( 1%, Fig. 2B). The 296
percentage of variance accounted for by the second eigenvector ( ) was significantly greater in 297
CA ( 0.002) indicating a wider gait loop (Fig. 2B). The orientation of the covariance plane ( 298
parameter) was also significantly different between the two groups ( 0.0002). 299
The stride-by-stride variability in gait kinematics was consistently larger in CA. Figure 3A 300
illustrates superimposed plots of the hip, knee, ankle, and trunk pitch and roll angles during 301
individual gait cycles in one control subject and one CA patient. On average, the inter-stride 302
variability in the angles, estimated as the mean SD over the gait cycle, was about 50% greater in 303
CA patients (Fig. 3B). 304
305
Ground reaction forces 306
Figure 4A illustrates the vertical component ( ) of the GRF for both individual strides in 307
single subjects (upper plots) and averaged curves for all subjects (lower plots). Both groups 308
showed the typical two-peaked profile, with a first maximum during the initial stance phase (∼25% 309
of stance phase) and the second one at the end of stance (∼80% of stance phase). A small 310
additional peak during weight acceptance, at ∼10% of stance, could also be observed (Fig. 4A), 311
consistent with previous studies in healthy subjects (Borghese et al., 1996; Verdini et al., 2006). 312
This peak was prominent in all CA patients, while it was seen only in a few control subjects at 313
these slow to moderate speeds. We quantified the heel strike transient by calculating the
314
parameter (Fig. 4A): was significantly larger ( 0.00004) in CA compared to HS (Fig. 4C). The 315
horizontal shear forces did not show systematic differences between the two groups and are not 316
reported. 317
We also examined whether the augmented transient in the vertical GRF in CA (Fig. 4A) 318
correlates with kinematic parameters. To this end, we computed the difference between maximum 319
and minimum values of the vertical velocity of the markers located on the hip, knee, and ankle, 320
measured at the beginning of stance (0-10% of the stance phase, in Fig. 4B). Examples of 321
velocity traces in one control subject and one CA patient are illustrated in Figure 4B. The 322
parameter was significantly larger ( 0.0001) for the knee and hip markers in patients (Fig. 4C), 323
consistent with a greater instability of limb loading in CA as shown by the prominent transient in 324
vertical force ( , Fig. 4A). 325
326
EMG envelopes 327
We recorded EMG signals from 12 muscles of the right lower limb. Figure 5A illustrates 328
EMG traces in one healthy subject and one CA patient during two consecutive strides. In both 329
groups, EMG activity during walking tended to occur in bursts that were temporally related to 330
specific kinematic/kinetic events: weight acceptance (VM, VL, RF, TFL, GM), limb 331
loading/propulsion (SOL, MG, LG, PL), foot lift (TA), heel strike (ST, BF). The normalized and 332
ensemble-averaged EMGs for the two groups of subjects are illustrated in Figure 5B. We noticed 333
that the EMG profiles in CA patients often differed relative to those in HS in two main features. 334
First, the major bursts tended to be wider (more prolonged) for most muscles. Second, EMG 335
profiles of CA could show some extra bursts, presumably related to gait instability (Fig. 5A). For 336
instance, the activity of hamstring muscles (BF and ST) in patients started earlier (at about 80% of 337
gait cycle) and was prolonged till about 50% of gait cycle with respect to the healthy subjects (Fig. 338
5B). A wider activity was also notable in the TA muscle. Similarly, calf muscles (SOL, MG, LG and 339
PL) in CA showed activity throughout the whole stance phase, even at the onset of stance, when 340
they are silent in HS. 341
EMG envelopes in Figure 5B were normalized to the maximum value across all trials, but 342
we also quantified the absolute mean activity and the extent of modulation of activity of leg 343
muscles over the gait cycle. Although large inter-individual variability was observed (in part due to 344
the individual differences in skin impedance), on average the mean and the max amplitude of 345
muscle activity over the gait cycle was about twice greater in CA patients ( 0.00001, Fig. 5C), 346
significantly increased activity in CA was found in both proximal and distal muscles. 347
CA patients showed significantly higher co-activation index values throughout the gait cycle 348
both for RF-VL-VM vs ST-BF (11.7±2.8 for CA and 9.3±2.1 for HS; 0.01) and MG-LG vs TA 349
(15.5±3.5 for CA and 8.7±2.1 for HS; 0.00001) pairs of antagonist muscles (Fig. 5D). 350
To characterize differences in timing and duration of EMG activity between HS and CA 351
groups, we computed the center of activity ( ) and full width at half maximum ( ) (Fig. 6A, 352
see Methods). The was similar for many muscles for the two groups of subjects, though it 353
shifted to slightly later phases of the gait cycle (CCW in the polar plots) in proximal muscles (VL, 354
ST, BF), and to earlier phases (CW in the polar plots) in distal muscles (SOL, MG, LG and PL) in 355
CA patients with respect to HS ( 0.00001, Fig. 6C). The analysis of allowed us to 356
quantify the duration of activity of each muscle. Figure 6B shows that most muscles (TA, 357
hamstrings and distal extensors) significantly increased their in CA patients in comparison 358
with HS. The mean (across all muscles) in HS and CA was 20% and 29% of the gait cycle, 359
respectively. Also, we verified whether the depended on velocity in the range of analyzed 360
walking speeds (3-4.5 km/h). The analysis did not reveal any significant correlation between 361
(expressed in % gait cycle) and walking speed in both groups of subjects (Fig. 6C), 362
consistent with scaling of EMG activity with cycle duration (Ivanenko et al., 2004; Cappellini et al., 363
2006). 364
Correlations between EMGs activation patterns, clinical scores and gait parameters in CA 366
Figure 7A illustrates significant correlations between muscle activation pattern 367
characteristics and kinematic and kinetic parameters in CA patients. We observed significant 368
relationships ( 0.02) between mean (averaged across all muscles) and cycle duration, 369
stride length, and stride width of CA patients (Fig. 7A). For the intersegmental coordination and 370
angular motion, we found a significant ( 0.01) correlation of only with (Fig. 7A). 371
Figure 7B, C and D show the relationship between clinical ICARS measures and gait 372
parameters (which were significantly different between HS and CA). The following parameters 373
correlated significantly with the ICARS score: cycle duration ( 0.04), stride length ( 0.04), 374
stride width ( 0.03), ( 0.02) (Fig. 7B), of the GRF ( 0.02), of the hip and knee 375
( 0.01 and 0.001, respectively) (Fig. 7C), and mean of EMG envelopes ( 0.002) 376
(Fig. 7D). The was similar among different forms of CA: 29.2±8.9% for SCA1, 28.9±7.1% 377
for SCA2 and 28.3±5.7% for SAOA. While there were significant differences in the interstride 378
variability in the kinematic parameters between CA and HS (Fig. 3), there was no simple 379
relationship between the increased stride-by-stride variability in CA patients and the clinical ICARS 380
measures (see also Ilg et al., 2007; Serrao et al., 2012). Thus, a large number of gait and muscle 381
pattern parameters that were significantly different between CA and HS gaits (Fig. 1, 2, 4, 6) 382
correlated with the severity of pathology (ICARS score). 383
Discussion 384
385
In this study we investigated muscle activity and the biomechanics of locomotion in a group 386
of patients diagnosed with SCA1, SCA2 and SAOA condition of CA. We analyzed the 387
characteristics of EMG activity of 12 unilateral lower limb muscles, correlating them with the clinical 388
score and global and segmental parameters extracted from the kinematics and ground reaction 389
forces. Our findings revealed new idiosyncratic features of the CA gait: significant changes in the 390
intersegmental coordination (Fig. 1C, 2), an abnormal transient in the vertical ground reaction force 391
and instability of limb loading at heel strike (Fig. 4). The marked feature of neuromuscular control 392
of gait in CA was the widening of EMG bursts (Fig. 5, 6). Below we discuss the relationship 393
between the primary deficits and/or compensatory strategies and adaptive changes in the walking 394
behavior and muscle activity patterns. 395
396
Kinematics features of CA gait 397
Several studies have compared the parameters characterizing locomotion of cerebellar 398
patients with that of healthy controls. The results confirmed high variability of spatiotemporal gait 399
parameters (Fig. 3), wide-base support and reduced cycle duration and stride length (Fig. 1A) that 400
has previously been shown to be distinctive features of ataxic locomotion aimed at compensating 401
the wide oscillations of the center of mass due to poor dynamic balance and stability (Palliyath et 402
al., 1998; Ilg et al., 2007; Serrao et al., 2012; Wuehr et al., 2013). Reduced RoM in the ankle joint 403
in CA could be related to shorter steps (Fig. 1A), impaired intersegmental coordination (wider gait 404
loop, Fig. 2) and/or stiffening of the ankle joint, which would reduce push-off forces, necessary to 405
propel the center of mass forward and upward, and thus destabilize the gait (Morton and Bastian, 406
2003). Increased trunk sway (Fig. 1C, right panel) can be assumed to be related to multidirectional 407
postural instability (Van de Warrenburg et al., 2005a). 408
The analysis of the intersegmental coordination in CA revealed significant changes in the 409
covariance plane orientation (expressed by , Fig. 2 right panel) and wider gait loop (expressed 410
by the higher values of PV2, Fig. 2). This is a significant finding, because the orientation of the
covariance plane reflects a specific tuning of the phase coupling between pairs of limb segments 412
and is related to an ability of the subject to adapt to different walking conditions (Bianchi et al., 413
1998; Dominici et al., 2010). For instance, in toddlers, the ability to adapt to different terrains is 414
very limited and the maintenance of a roughly constant planar covariance reduces flexibility of the 415
kinematic pattern and thus restricts the manifold of angular segment motion (Dominici et al., 2010). 416
Likewise, it would be of a great interest to study whether there is also a lack of specific adaptation 417
in the planar covariation of limb segments in CA during walking over different terrains (see, for 418
instance, MacLellan et al., 2011). 419
420
Foot loading in CA 421
In cerebellar patients, the vertical ground reaction force demonstrated a prominent transient 422
following the heel strike (Fig. 4), indicating an abnormal control of limb loading, correlated with the 423
severity of the pathology (Fig. 7C). Even though it can be observed in healthy individuals during 424
fast walking (Borghese et al., 1996), normally the intensity of this impact is adjusted by shock-425
absorbing reactions at the ankle, knee and hip (Perry, 1992) and it is small or absent at low and 426
normal walking speeds (Fig. 4A, left panels). A similar peak was also found in other pathologies, 427
like osteoarthritis and low back pain (Collins and Whittle, 1989), in amputees (Klodd et al., 2010), 428
and in healthy subjects during unstable walking on a slippery surface (Cappellini et al., 2009). 429
Despite its deceiving simplicity, human locomotion incorporating an heel strike and 430
appropriate heel-to-toe rolling pattern during stance is a precise and complex motor task that 431
requires learning (Dominici et al., 2007; Ivanenko et al., 2007). The cause of the impaired loading 432
in CA (Fig. 4) may originate from the unbalanced control and preparation to the foot touchdown. An 433
increase in the impact transient could also be a consequence of leg stiffening in CA patients (Mari 434
et al., 2014) or reduced push-off of the contralateral limb in late stance (Serrao et al., 2012). For 435
instance, in amputees, the occurrence of the heel strike transient is evident on the sound side 436
while the prosthetic limb exhibits smooth loading, presumably due to a lack of active push-off from 437
the prosthetic feet in late stance and/or reduced energy storage and return from the prosthetic feet 438
(Klodd et al., 2010). Nevertheless, the weakness of distal extensors does not inevitably result in 439
the abnormal GRF transient since it was not observed in peripheral neuropathy (Ivanenko et al., 440
2013a). Further experiments are needed to understand better its biomechanical nature. Whatever 441
the exact biomechanical reasons for the observed phenomenon, the cerebellum may play an 442
important role in the foot loading control. For instance, cats with unilateral section of the dorsal 443
spinocerebellar tract cannot walk on a slippery floor (R.E. Poppele, unpublished observation) as 444
well as cerebellar gait ataxia in humans may result in leg-placement deficit (Morton and Bastian, 445
2003). 446
447
Muscle activation patterns in cerebellar ataxia 448
Despite that the kinematics and bilateral coordination of leg muscle activity is quite 449
symmetrical in normal healthy subjects, in pathological conditions there might be some differences 450
(Perry, 1992). We did not find any significant difference in the kinematic parameters and clinical 451
assessment on both sides in CA patients. However, although no relevant clinical and kinematic 452
asymmetry were found in our patients we cannot exclude subclinical differences in the EMG 453
patterns between left and right sides. The recordings of unilateral muscle activity, nevertheless, 454
revealed distinctive features of EMG bursts in CA with respect to HS (Fig. 5, 6, 7D). 455
Although the muscular patterns of CA patients are variable, the analysis of muscle 456
activation patterns showed distinctive features of the CA gait, in particular an increased amplitude 457
(Fig. 5C) and duration (Fig. 6B) of EMG bursts. On average the amplitude of muscle activity over 458
the gait cycle was about twice in CA patients than in healthy subjects (Fig. 5C, see also Mitoma et 459
al., 2000). It is therefore remarkable that significantly larger muscle activation could lead to similar 460
leg movement kinematics and even to smaller angular oscillations in the ankle joint (Fig. 1B, 2A). 461
In addition to a relatively high level of muscle activity in patients (Fig. 5C), the main difference 462
between HS and CA was the duration of the muscle activation periods (Fig. 6B). The enlarged 463
was observed in all three groups of CA patients (SCA1, SCA2, SAOA). The widening of 464
EMG bursts was somewhat asymmetric since there were also changes in the (Fig. 6C): the 465
shifted to slightly later phases of the gait cycle in proximal muscles (VL, ST, BF), and to earlier 466
phases in distal muscles (SOL, MG, LG and PL). 467
The increased co-activation observed in the cerebellar ataxia patients (Fig. 6D, see also 468
Mari et al., 2014), in part due to EMG widening (Fig. 5), may represent a compensatory strategy 469
useful to provide mechanical stability by stiffening joints. For instance, an abnormal co-contraction 470
pattern has been demonstrated in categories of people who have a great need for active muscular 471
stabilization, such as the elderly (Peterson and Martin, 2010), individuals who have undergone 472
knee arthroplasty (Fallah-Yakhdani et al., 2012), patients with stroke or traumatic brain injury 473
(Chow et al., 2012), and patients with Parkinson’s disease (Meunier et al., 2000). Indeed, 474
compared with healthy subjects, ataxic patients needed to activate antagonist muscles more and 475
for a longer period, possibly in an attempt to compensate for instability (Fig. 1C, 3, 4). The 476
observed EMG widening also correlated with a stereotyped biomechanical output of the CA gait 477
(Fig. 7A). However, while leg stiffening might be beneficial in reducing body oscillations during 478
normal posture, in dynamic conditions it may also be detrimental due to a complex nature of 479
balance control during walking. Therefore, an alternative explanation could be that the broader 480
activity bursts are a result of pathology, as suggested by the positive correlation between the 481
severity of pathology (clinical ataxia scale, ICARS) and the (Fig. 7D). Nevertheless, taking 482
into account the ability of the central nervous system to adapt when faced with a specific gait 483
pathology, often it is difficult to distinguish what primarily comes from pathology and what comes 484
from compensatory mechanisms (Dietz, 2002; Grasso et al., 2004; Ivanenko et al., 2013a). The 485
observed widening of EMG bursts (Fig. 6B) can possibly be compared to other gait disturbances or 486
gait adaptations. For instance, relatively wider EMG bursts are observed in infants (Dominici et al., 487
2011; Ivanenko et al., 2013b), which may be determined at least in part by the developmental state 488
of the cerebellum (Vasudevan et al., 2011). Similarly to ataxic patients, when children start to walk 489
independently their gait is characterized by considerable trunk oscillations, wide swinging arms, 490
high interstride variability and immature foot trajectory characteristics and intersegmental 491
coordination (Ivanenko et al., 2007; Dominici et al., 2010). Maturation of gait is accompanied by a 492
more selective and flexible control of muscles, with shorter activations and an evident separation of 493
the distinct bursts (Dominici et al., 2011; Teulier et al., 2012; Ivanenko et al., 2013b). 494
What is the advantage of the ‘narrow’ bursts in the activation patterns of HS and why are 495
broader bursts adopted in CA patients? Even though the central pattern generation ‘timer’ 496
produces different relative stance/swing phase durations depending on walking speed (Prochazka 497
and Yakovenko, 2007; Duysens et al., 2013), the duration of the muscle activation patterns is 498
scaled to the duration of the gait cycle (Cappellini et al., 2006). In cerebellar ataxia patients, 499
widening of muscle activation patterns and shifts in the (Fig. 6) may be a consequence of 500
improper motor planning (feed-forward control) and processing of proprioceptive information 501
(Bastian, 2011) leading to inaccurate movements (Fig. 3) and to the abnormal transient at heel 502
strike (Fig. 4). Broader activation patterns likely imply higher metabolic cost and may also limit 503
adaptation to different walking conditions and coordination with voluntary movements that require 504
appropriate activation timings/duration (Ivanenko et al., 2005). 505
Our findings are consistent with the idea that the cerebellum contributes to optimizing the 506
duration of muscle activation patterns during locomotion. It remains to be determined if the 507
abnormalities discussed here are specific for cerebellar deficit. In this regard, it is worth noting that 508
abnormal prolongation of EMG activity was also observed in the upper limb muscles during elbow 509
flexions in patients with cerebellar deficits and thus may represent a general feature of cerebellar 510
dysmetria (Hallett et al., 1975). Future research can be focused on the mechanisms underlying the 511
observed phenomena for understanding cerebellar physiology and for using these abnormalities as 512
diagnostic tools for the documentation of cerebellar deficits. 513
Acknowledgements: This work was supported by the Italian Health Ministry, Italian Ministry of 515
University and Research (PRIN project), Italian Space Agency (CRUSOE and COREA grants), and 516
European Union FP7-ICT program (AMARSi grant #248311). 517
References 518
519
Barbero M, Merletti R, Rainoldi A. Atlas of Muscle Innervation Zones: Understanding Surface 520
Electromyography and Its Applications. Springer, 2012. 521
Bastian AJ. Moving, sensing and learning with cerebellar damage. Curr Opin Neurobiol 21: 596– 522
601, 2011. 523
Batschelet E. Circular Statistics in Biology. Academic Press, 1981. 524
Bianchi L, Angelini D, Orani GP, Lacquaniti F. Kinematic coordination in human gait: relation to 525
mechanical energy cost. J Neurophysiol 79: 2155–2170, 1998. 526
Borghese NA, Bianchi L, Lacquaniti F. Kinematic determinants of human locomotion. J Physiol 527
494: 863–879, 1996. 528
Bosco G, Eian J, Poppele RE. Phase-specific sensory representations in spinocerebellar activity 529
during stepping: evidence for a hybrid kinematic/kinetic framework. Exp Brain Res 175: 83–96, 530
2006. 531
Cappellini G, Ivanenko YP, Dominici N, Poppele RE, Lacquaniti F. Motor Patterns During 532
Walking on a Slippery Walkway. J Neurophysiol 103: 746–760, 2009. 533
Cappellini G, Ivanenko YP, Poppele RE, Lacquaniti F. Motor patterns in human walking and 534
running. J Neurophysiol 95: 3426–3437, 2006. 535
Chow JW, Yablon SA, Stokic DS. Coactivation of ankle muscles during stance phase of gait in 536
patients with lower limb hypertonia after acquired brain injury. Clin Neurophysiol Off J Int Fed Clin 537
Neurophysiol 123: 1599–1605, 2012. 538
Collins JJ, Whittle MW. Impulsive forces during walking and their clinical implications. Clin 539
Biomech Bristol Avon 4: 179–187, 1989. 540
Davis III RB, Õunpuu S, Tyburski D, Gage JR. A gait analysis data collection and reduction 541
technique. Hum Mov Sci 10: 575–587, 1991. 542
DeMyer W. Technique of the Neurologic Examination, Fifth Edition. McGraw Hill Professional, 543
2003. 544
Dietz V. Proprioception and locomotor disorders. Nat Rev Neurosci 3: 781–790, 2002. 545
Dominici N, Ivanenko YP, Cappellini G, d’ Avella A, Mondì V, Cicchese M, Fabiano A, Silei T, 546
Di Paolo A, Giannini C, Poppele RE, Lacquaniti F. Locomotor primitives in newborn babies and 547
their development. Science 334: 997–999, 2011. 548
Dominici N, Ivanenko YP, Cappellini G, Zampagni ML, Lacquaniti F. Kinematic strategies in 549
newly walking toddlers stepping over different support surfaces. J Neurophysiol 103: 1673–1684, 550
2010. 551
Dominici N, Ivanenko YP, Lacquaniti F. Control of foot trajectory in walking toddlers: adaptation 552
to load changes. J Neurophysiol 97: 2790–2801, 2007. 553
Duysens J, De Groote F, Jonkers I. The flexion synergy, mother of all synergies and father of 554
new models of gait. Front Comput Neurosci 7: 14, 2013. 555
Earhart GM, Bastian AJ. Selection and coordination of human locomotor forms following 556
cerebellar damage. J Neurophysiol 85: 759–769, 2001. 557
Fallah-Yakhdani HR, Abbasi-Bafghi H, Meijer OG, Bruijn SM, van den Dikkenberg N, 558
Benedetti M-G, van Dieën JH. Determinants of co-contraction during walking before and after 559
arthroplasty for knee osteoarthritis. Clin Biomech Bristol Avon 27: 485–494, 2012. 560
Goodworth AD, Paquette C, Jones GM, Block EW, Fletcher WA, Hu B, Horak FB. Linear and 561
angular control of circular walking in healthy older adults and subjects with cerebellar ataxia. Exp 562
Brain Res 219: 151–161, 2012. 563
Grasso R, Ivanenko YP, Zago M, Molinari M, Scivoletto G, Castellano V, Macellari V, 564
Lacquaniti F. Distributed plasticity of locomotor pattern generators in spinal cord injured patients. 565
Brain J Neurol 127: 1019–1034, 2004. 566
Grasso R, Peppe A, Stratta F, Angelini D, Zago M, Stanzione P, Lacquaniti F. Basal ganglia 567
and gait control: apomorphine administration and internal pallidum stimulation in Parkinson’s 568
disease. Exp Brain Res 126: 139–148, 1999. 569
Grillner S, Deliagina T, Ekeberg O, el Manira A, Hill RH, Lansner A, Orlovsky GN, Wallén P. 570
Neural networks that co-ordinate locomotion and body orientation in lamprey. Trends Neurosci 18: 571
270–279, 1995. 572
Hallett M, Shahani BT, Young RR. EMG analysis of patients with cerebellar deficits. J Neurol 573
Neurosurg Psychiatry 38: 1163–1169, 1975. 574
Hermens HJ, Freriks B, Merletti R, Stegeman D, Blok J, Rau G, Disselhorst-Klug C, Hägg G. 575
European recommendations for surface electromyography [Online]. Roessingh Research and 576
Development The Netherlands. http://www.seniam.org/pdf/contents8.PDF [6 Feb. 2014]. 577
Holmes G. The Cerebellum of Man. Brain 62: 1–30, 1939. 578
Horak FB, Diener HC. Cerebellar control of postural scaling and central set in stance. J 579
Neurophysiol 72: 479–493, 1994. 580
Ilg W, Golla H, Thier P, Giese MA. Specific influences of cerebellar dysfunctions on gait. Brain J 581
Neurol 130: 786–798, 2007. 582
Ilg W, Timmann D. Gait ataxia—specific cerebellar influences and their rehabilitation. Mov Disord 583
28: 1566–1575, 2013. 584
Ivanenko YP, d’ Avella A, Poppele RE, Lacquaniti F. On the origin of planar covariation of 585
elevation angles during human locomotion. J Neurophysiol 99: 1890–1898, 2008. 586
Ivanenko YP, Cappellini G, Dominici N, Poppele RE, Lacquaniti F. Coordination of locomotion 587
with voluntary movements in humans. J Neurosci Off J Soc Neurosci 25: 7238–7253, 2005. 588
Ivanenko YP, Cappellini G, Solopova IA, Grishin AA, MacLellan MJ, Poppele RE, Lacquaniti 589
F. Plasticity and modular control of locomotor patterns in neurological disorders with motor deficits. 590
Front Comput Neurosci 7, 2013a. 591
Ivanenko YP, Dominici N, Cappellini G, Paolo AD, Giannini C, Poppele RE, Lacquaniti F. 592
Changes in the Spinal Segmental Motor Output for Stepping during Development from Infant to 593
Adult. J Neurosci 33: 3025–3036, 2013b. 594
Ivanenko YP, Dominici N, Lacquaniti F. Development of independent walking in toddlers. Exerc 595
Sport Sci Rev 35: 67–73, 2007. 596
Ivanenko YP, Poppele RE, Lacquaniti F. Five basic muscle activation patterns account for 597
muscle activity during human locomotion. J Physiol 556: 267–282, 2004. 598
Klodd E, Hansen A, Fatone S, Edwards M. Effects of prosthetic foot forefoot flexibility on gait of 599
unilateral transtibial prosthesis users. J Rehabil Res Dev 47: 899–910, 2010. 600
Konczak J, Timmann D. The effect of damage to the cerebellum on sensorimotor and cognitive 601
function in children and adolescents. Neurosci Biobehav Rev 31: 1101–1113, 2007. 602
Lacquaniti F, Ivanenko YP, Zago M. Kinematic control of walking. Arch Ital Biol 140: 263–272, 603
2002. 604
Lennard PR, Stein PS. Swimming movements elicited by electrical stimulation of turtle spinal 605
cord. I. Low-spinal and intact preparations. J Neurophysiol 40: 768–778, 1977. 606
Leurs F, Bengoetxea A, Cebolla AM, De Saedeleer C, Dan B, Cheron G. Planar covariation of 607
elevation angles in prosthetic gait. Gait Posture 35: 647–652, 2012. 608
MacLellan MJ, Dupré N, McFadyen BJ. Increased obstacle clearance in people with ARCA-1 609
results in part from voluntary coordination changes between the thigh and shank segments. 610
Cerebellum Lond Engl 10: 732–744, 2011. 611
Mari S, Serrao M, Casali C, Conte C, Martino G, Ranavolo A, Coppola G, Draicchio F, Padua 612
L, Sandrini G, Pierelli F. Lower limb antagonist muscle co-activation and its relationship with gait 613
parameters in cerebellar ataxia. Cerebellum Lond Engl 13: 226–236, 2014. 614
Meunier S, Pol S, Houeto JL, Vidailhet M. Abnormal reciprocal inhibition between antagonist 615
muscles in Parkinson’s disease. Brain J Neurol 123 ( Pt 5): 1017–1026, 2000. 616
Mitoma H, Hayashi R, Yanagisawa N, Tsukagoshi H. Characteristics of parkinsonian and ataxic 617
gaits: a study using surface electromyograms, angular displacements and floor reaction forces. J 618
Neurol Sci 174: 22–39, 2000. 619
Morton SM, Bastian AJ. Relative Contributions of Balance and Voluntary Leg-Coordination 620
Deficits to Cerebellar Gait Ataxia. J Neurophysiol 89: 1844–1856, 2003. 621
Morton SM, Bastian AJ. Cerebellar control of balance and locomotion. Neurosci Rev J Bringing 622
Neurobiol Neurol Psychiatry 10: 247–259, 2004. 623
Morton SM, Bastian AJ. Mechanisms of cerebellar gait ataxia. Cerebellum Lond Engl 6: 79–86, 624
2007. 625
Nardone A, Schieppati M. The role of instrumental assessment of balance in clinical decision 626
making. Eur J Phys Rehabil Med 46: 221–237, 2010. 627
Palliyath S, Hallett M, Thomas SL, Lebiedowska MK. Gait in patients with cerebellar ataxia. Mov 628
Disord Off J Mov Disord Soc 13: 958–964, 1998. 629
Perry J. Gait analysis: normal and pathological function. Thorofare, NJ: SLACK, 1992. 630
Peterson DS, Martin PE. Effects of age and walking speed on coactivation and cost of walking in 631
healthy adults. Gait Posture 31: 355–359, 2010. 632
Prochazka A, Yakovenko S. The neuromechanical tuning hypothesis. Prog Brain Res 165: 255– 633
265, 2007. 634
Roberts A, Tunstall MJ, Wolf E. Properties of networks controlling locomotion and significance of 635
voltage dependency of NMDA channels: stimulation study of rhythm generation sustained by 636
positive feedback. J Neurophysiol 73: 485–495, 1995. 637
Rossignol S, Chau C, Brustein E, Giroux N, Bouyer L, Barbeau H, Reader TA. 638
Pharmacological activation and modulation of the central pattern generator for locomotion in the 639
cat. Ann N Y Acad Sci 860: 346–359, 1998. 640
Rudolph KS, Axe MJ, Snyder-Mackler L. Dynamic stability after ACL injury: who can hop? Knee 641
Surg Sports Traumatol Arthrosc Off J ESSKA 8: 262–269, 2000. 642
Serrao M, Pierelli F, Ranavolo A, Draicchio F, Conte C, Don R, Di Fabio R, LeRose M, Padua 643
L, Sandrini G, Casali C. Gait pattern in inherited cerebellar ataxias. Cerebellum Lond Engl 11: 644
194–211, 2012. 645
Stolze H, Klebe S, Petersen G, Raethjen J, Wenzelburger R, Witt K, Deuschl G. Typical 646
features of cerebellar ataxic gait. J Neurol Neurosurg Psychiatry 73: 310–312, 2002. 647
Teulier C, Sansom JK, Muraszko K, Ulrich BD. Longitudinal changes in muscle activity during 648
infants’ treadmill stepping. J Neurophysiol 108: 853–862, 2012. 649
Thach WT, Bastian AJ. Role of the cerebellum in the control and adaptation of gait in health and 650
disease. Prog Brain Res 143: 353–366, 2004. 651
Trouillas P, Takayanagi T, Hallett M, Currier RD, Subramony SH, Wessel K, Bryer A, Diener 652
HC, Massaquoi S, Gomez CM, Coutinho P, Ben Hamida M, Campanella G, Filla A, Schut L, 653
Timann D, Honnorat J, Nighoghossian N, Manyam B. International Cooperative Ataxia Rating 654
Scale for pharmacological assessment of the cerebellar syndrome. The Ataxia 655
Neuropharmacology Committee of the World Federation of Neurology. J Neurol Sci 145: 205–211, 656
1997. 657
Vasudevan EVL, Torres-Oviedo G, Morton SM, Yang JF, Bastian AJ. Younger Is Not Always 658
Better: Development of Locomotor Adaptation from Childhood to Adulthood. J Neurosci 31: 3055– 659
3065, 2011. 660
Verdini F, Marcucci M, Benedetti MG, Leo T. Identification and characterisation of heel strike 661
transient. Gait Posture 24: 77–84, 2006. 662
Van de Warrenburg BPC, Bakker M, Kremer BPH, Bloem BR, Allum JHJ. Trunk sway in 663
patients with spinocerebellar ataxia. Mov Disord Off J Mov Disord Soc 20: 1006–1013, 2005a. 664
Van de Warrenburg BPC, Steijns JAG, Munneke M, Kremer BPH, Bloem BR. Falls in 665
degenerative cerebellar ataxias. Mov Disord Off J Mov Disord Soc 20: 497–500, 2005b. 666
Watson GS, Williams EJ. On the Construction of Significance Tests on the Circle and the Sphere. 667
Biometrika 43: 344, 1956. 668
Winter DA. Biomechanics of human movement. Wiley, 1979. 669
Winter DA. The biomechanics and motor control of human gait: normal, elderly and pathological. 670
University of Waterloo Press, 1991. 671
Wuehr M, Schniepp R, Ilmberger J, Brandt T, Jahn K. Speed-dependent temporospatial gait 672
variability and long-range correlations in cerebellar ataxia. Gait Posture 37: 214–218, 2013. 673
Figure legends 675
676
Figure 1. Kinematics gait parameters. A: comparison of general gait parameters for healthy 677
subjects (HS) and cerebellar ataxic patients (CA). B: ensemble-averaged (mean ± SD) hip, knee 678
and ankle joint angles and trunk roll and pitch orientation angles in 19 ataxic patients (right) and 20 679
age-matched healthy subjects (left). Data were normalized to the cycle duration and represented in 680
percent of gait cycle. C: range of angular motion (RoM). Asterisks denote significant group 681
differences ( 0.05, un-paired t-tests). 682
683
Figure 2. Planar covariation of elevation angles during walking in HS and CA. A: ensemble-684
averaged (across strides) thigh, shank and foot elevation angles (mean ± SD) plotted vs. 685
normalized gait cycle and corresponding 3D gait loops and interpolation planes in one healthy 686
subject (left) and one ataxic patient (right). Gait loops are obtained by plotting the thigh waveform 687
vs. the shank and foot waveforms (after mean values subtraction). Gait cycle paths progress in 688
time in the counterclockwise direction, heel touch-down and toe-off phases corresponding roughly 689
to the top and bottom of the loops, respectively. The interpolation planes result from orthogonal 690
planar regression. B: percent of total variation explained by second and third principal component 691
( and , respectively) and parameter that characterizes the orientation of the normal to 692
the plane are indicated for each group of subjects (mean + SD). Asterisks represent significant 693
group difference ( 0.05). 694
695
Figure 3. Inter-stride angular variability. A: examples of joint and trunk orientation angles in CA 696
patient (p14) and an age-matched control (h1). Every trace refers to a single cycle. B: inter-stride 697
variability of joint and trunk orientation angles (meanSD: estimated as the mean SD of angular 698
waveforms across strides averaged over all time points of the gait cycle) in healthy subjects and 699
ataxic patients (mean + SD). Asterisks represent significant group difference ( 0.05). Note 700
higher inter-stride variability of angular motion in CA. 701
Figure 4. Vertical ground reaction forces (GRF) during walking. A: upper panels: vertical GRF in 703
one representative healthy subject (top left) and one ataxic patient (top right). Every trace refers to 704
a single cycle. Bottom panels: Ensemble-averaged (mean ± SD) vertical ground reaction forces in 705
healthy subjects ( 20) and ataxic patients ( 19). The patterns are normalized to body weight 706
and plotted vs. normalized stance. Note the prominent transient ( ) during the weight-acceptance 707
period (marked by a shaded area) in ataxic patients. B: vertical velocity of the markers at hip, knee 708
and ankle joints in one healthy subject (left) and one representative ataxic patient (right) over the 709
time interval around the foot-strike event (from 5% of stance prior to the heel contact to 15% after 710
the heel contact). Each trace refers to a single step. – peak-to-peak amplitude of velocity traces 711
over first 10% of stance duration. C: peak-to-peak amplitude (mean + SD) of force transient ( ) 712
and velocity traces ( ) during the initial weight acceptance phase of stance in healthy subjects 713
and ataxic patients. Asterisks denote significant group differences ( 0.05). 714
715
Figure 5. EMG activity in HS and CA. A: examples of EMG traces in one healthy subject (h4, 716
4.1km/h) and one CA patient (p15, 3.4km/h) during two consecutive strides. The stance phase is 717
evidenced by a shaded region. B: ensemble-averaged (mean ± SD) EMG activity patterns of 12 718
ipsilateral leg muscles recorded from healthy subjects and ataxic patients. EMGs were normalized 719
to their max value across all trials. C: max and mean EMG levels (mean + SD) in microvolts. Note 720
higher level of activity in CA. D: co-activation indexes ( ) of “RF-VL-VM vs ST-BF” and “MG-LG vs 721
TA” pairs of antagonist muscles. Asterisks denote significant group differences ( 0.05). 722
723
Figure 6. Characteristics of EMG activity. A: schematic description of the evaluation method in one 724
representative EMG envelope of the biceps femoris muscle plotted in the polar coordinates. Full 725
width at half maximum ( ) was calculated as the duration of the interval (in percent of gait 726
cycle) in which EMG activity exceeded half of its maximum. In the few cases in which two bursts of 727
activity were present (e.g. in TA), was calculated as a sum of the durations of the intervals 728
in which EMG activity exceeded half of its maximum. The center of activity ( ) vector was 729
calculated as the first trigonometric moment of the circular distribution (Batschelet, 1981). B: 730
(mean + SD) of 12 EMGs in HS and CA. C: correlation between and walking 731
speed, the data for all subjects and all individual strides were pooled together (each point 732
represents the individual stride value). Linear regression lines with corresponding and values 733
are reported. D: of leg muscle EMGs in healthy subjects (green) and ataxic patients (red). 734
Polar direction denotes the relative time over the gait cycle (time progresses clockwise), the width 735
of the sector represents angular SD across subjects while the radius of the sector indicates the 736
mean angular SD across strides (the smaller the radius the larger the interstride variability). 737
Asterisks denote significant differences between the groups. 738
739
Figure 7. Correlations between gait parameters, EMG burst widening and clinical scores. Only 740
parameters that differed significantly between HS and CA individuals are plotted in this figure. 741
Each point represents the stride-averaged value for the individual patient. Linear regression lines 742
with corresponding and values are reported. A: Relationships between cycle duration, stride 743
length, stride width and parameter of the intersegmental coordination and mean of 744
muscle activation patterns (mean was calculated as the mean across of all 12 745
muscles). B: gait kinematic parameters vs. the ICARS score. C: parameters of the transient 746
following the heel strike (GRF , hip and knee , Fig. 4). D: averaged across all 747
EMGs vs. the ICARS score. 748
Table 1. Patients’ characteristics. Cerebellar patients were rated using the ICARS score (Trouillas 749
et al., 1997). The table lists the total ICARS scores and the subscores for posture, gait and limb 750
kinetics (we summed up the gait and posture scores to obtain an indicator of balance deficit). 751
Higher scores indicate more severe ataxia. 752
753
Patients Age (yr) Gender BW (kg) Diagnosis onset (yr) Age at ICARS
Gait Posture Balance Lower Limb Total
P1 42 M 65 SCA1 35 1 1 2 1 6 P2 41 M 64 SCA1 33 1 1 2 1 6 P3 48 M 74 SAOA 30 1 0 1 1 6 P4 32 M 50 SCA1 28 2 4 6 0 7 P5 33 M 52 SCA1 30 3 3 6 0 7 P6 57 M 65 SAOA 47 3 4 7 4 12 P7 65 F 65 SAOA 62 3 5 8 0 12 P8 59 F 61 SAOA 55 3 4 7 1 12 P9 43 F 66 SCA2 37 5 5 10 2 17 P10 49 M 73 SCA1 41 4 5 9 2 18 P11 46 M 77 SAOA 17 4 5 9 2 18 P12 37 M 67 SCA2 30 3 6 9 2 20 P13 45 M 79 SCA1 27 3 3 6 3 21 P14 45 M 70 SCA2 35 4 8 12 5 21 P15 44 M 68 SCA2 33 5 7 12 5 21 P16 52 M 85 SCA1 40 4 5 8 4 23 P17 45 M 68 SAOA 30 4 9 13 7 26 P18 62 F 70 SAOA 50 5 9 14 7 28 P19 54 F 66 SAOA 40 5 10 15 8 30
A
2 4 km /h stance duration 100 40 60 80 stride length 0 5 1.0 1.5 limb length * stride width 0.2 0.4 limb length * s cycle duration 0.5 1.0 * walking speed % cy cle 0 20 0B
healthy subjects CA patients 0° 30° hip 0 0.5 % 0 % 0 -20° 0° 80° 80° knee kl 40° 60° -5° 0° 5° ankle trunk roll 0 50 100 -5° 0° 5° % cycle 0 50 100 trunk pitch % cycle R MC
RoMC
HS CAhip knee ankle 0° 20° 40° 60° * roll pitch 0° 2° 4° 6° * * trunk
Fig. 1
A
healthy h4 patient p12 foot thigh shank 60° 0° 60° ootB
PV3 * PV2 * u*3t -60° 0 fo -60° 60° -60° 0° 60° 0° HS 0% 0.5% 1% 0% 5% 10% 15% 0 0.1 0.2 0.3 HS CAFig. 2
healthy h1 patient p14 60° 80° 80° 100°
A
40° 60° 0° 20° 40° 60° 60° 80° 0° 20° 40° 60° ankle knee 0 -40° -20° 0° 20° 10° 0 -40° -20° 0° 20° 10° hip -10° 0° 0° 5° -10° 0° 0° 5° trunk roll trunk pitch 0 50 100 -5° 0 50 100 -5° % cycle % cycleB
S D SD * 6° 8° * 3° 4° * * * inter-stride variability HS CAankle knee hip
mean S roll pitch mean S 0° 2° 4° 0° 1° 2° trunk CA
Fig. 3
A
vertical GRF healthy h1 patient p14B
vertical velocity 5 healthy h1 patient p14 0 0.2 0.4 0.6 0.8 1 % BW Δ1 -5 0 5 m/ s 5 -5 0 m/ s Δ2 knee hip 0.4 0.6 0.8 1 % BWhealthy subjects CA patients
C
-5 0 5 10 15 -5 0 5 % stance m/ s -5 0 5 10 15 % stance 6 * * 0.2 vert. GRF * ankle vert. velocity HS CA 0 20 40 60 80 100 0 0.2 % stance 0 20 40 60 80 100 % stance hip knee ankle 0 2 4 Δ2 0 0.05 0.1 0.15 Δ1Fig. 4
B
healthy subjects CA patients healthy h4 patient p15A
0 1 RF VL RF VL 0 1 VM GM TFL VM GM TFL ST BF ST BF TA PL TA PL SOL MG LG SOL MG LGC
mean activation μ V40 60 * max activation 400 600 * μ V 1 s 1 s HS CA 10% 20%co-activation index (CI)
* * HS CA
D
0 50 100 % cycle 0 50 100 % cycleFig. 5
0 20 0 200 μ RF-VL-VM vs. ST-BF MG-LG vs. TA 0%A
FWHM 40 60 cle * * * * * * *B
0 FWHM CoA HS CA 30 40 50 rCA = -0.26, p=0.27 % cy cle) HS CAC
RF VL VM GM TFL STcenter of activity (CoA)
